DESCRIPTION: The study will use mammography films randomly sampled from a typical screening population with disease status determined before sampling by biopsy or long-term follow up and subsequent negative mammography exam. These films will be interpreted by radiologists from randomly sampled facilities having a mammography unit accredited by the American College of Radiology. Regression models for ROC curves will be used to relate variability in interpretation and diagnostic accuracy as functions of radiologist and patient characteristics. The influence of comparison exams, geographic variability and intra-reader variability on mammogram interpretation will be investigated.